NHL.com will start tracking "advanced" stats

[mrkajz44]

Originally Posted by Street Pharmacist

Some Flyers fans did it too. In the end, why does it matter if it measures "possession"? It's a red herring. It's usefulness and predictive power aren't related to whether or not it measures "possession". Measuring who throws the puck towards the opposition's net more often seems to me to be more important in determining who played better over who had the puck more.

So the test to see if corsi is a good proxy for possession has only been done twice?

I do see what you are saying that it doesn't matter if corsi is a good proxy for possession or not, just that corsi seems to correlate with winning. It's just that any advanced stats articles seems to draw the conclusion that corsi = possession, and I was wondering about the support for that truism.

[CorsiHockeyLeague]

Well, first of all, intuitively it's actually a proxy for offensive zone possession. In order to direct a shot attempt at your opponent's goal, you must have the puck in the offensive zone. For the same reason, goals or shots would also be a proxy for offensive zone possession, but they're significantly rarer than shot attempts. So corsi and fenwick are used essentially because of sample size. For every game, teams each take an average of 29 or so shots. However, they direct more than twice as many shot attempts at the net. So you have twice as many instances where you can say, "this team had the puck in the offensive zone".

Does it, practically speaking, correlate to offensive puck possession in the sense of "guy on the Canadiens had the puck on his stick in the offensive zone for X second"? All attempts to track this have suggested it's very close.


Because of the time and effort it takes to hold two stopwatches in hand, it's not done very much. There may be other instances of people doing this, but once you've done it for a bit and see "yeah, it's basically the same", I would imagine your desire to keep going wanes. There's never been anyone I know of who's attempted this and found a significant difference between corsi and actual time of possession. We'll know for certain once sportVU comes in, though. It would be very surprising to see anything significantly different from that graph.

[nfotiu]

Originally Posted by Tinordi

You should only care about that stat if it correlates to something that you care about. If time of possession has no correlation to winning then there's no point in tracking the stat. Which is why so many people are focusing on corsi close, because it correlates to winning, meaning that it is more or less predictive in its value.

Is there a link to something that backs that up? I'd be interested in seeing how corsi correlates to winning vs regular shot counts vs time of possession.

I'd also be interested in seeing if factors like having an elite goaltender, defenceman, play-maker or goal scorer can allow a team to outperform those stats in the long term. Or if different team systems and styles can also allow a team to under or over perform the advanced stats.

Baseball stats have matured to that point and have become somewhat reliable predictors of success. Maybe I'm just not looking in the right places, but I haven't seen anything that convinces me that corsi and pdo are great predictors.

[CorsiHockeyLeague]

Originally Posted by nfotiu

Is there a link to something that backs that up? I'd be interested in seeing how corsi correlates to winning vs regular shot counts vs time of possession

http://www.pensionplanpuppets.com/20...-blocked-shots

Shows the correlation and realiability with respect to winning of lots and lots of different things.

As for outperforming them in the long-term, there are essentially three major areas.

1. Goaltending. This is sustainable long-term in the sense that some goalies are just better than league average. Hence, the Canadiens will never fare as badly as their ES possession says they should.
2. Special Teams. Possession is an even strength stat, so if your special teams goal differential is huge, well, you get the idea.
3. Finishing ability. This is reflected in team shooting percentage. If it's significantly above 8%, then you may be outscoring some of your other shortcomings. The Leafs have traditionally had a higher sh% that buoyed some bad possesesion teams for a while. The problem is, history has shown it's not as sustainable as the other two, particularly goaltending.